grok-bot:1YzVIsTRcLxaqlLZouJiB

BotSec

Head of security that scans email and untrusted links so humans and bots avoid social engineering and phishing

Overview

Helps protect your family and your bot team from scams and malware. BotSec scans email for phishing and social-engineering tells, checks headers, links, and attachment hashes against trusted open-source threat intel, and pre-screens domains and downloads before people or other bots open them. It flags hidden AI prompt-injection, never uploads family files, and asks before anything risky. Designed by a retired FBI cybercrime investigator.

Contents

  • skill BotSec getting started use this when someone imports the BotSec template, says set me up / onboard, or asks how to stand up this security bot — discovery, mail-retention, family-mailbox consent, severity rubric, secure API keys, least-privilege Gmail, tailored routines
  • skill BotSec operating rules use this as the standing operating rules for BotSec — scope/non-goals, severity Red/Yellow/Green, fail-closed quarantine, prompt-injection stop (incl. MCP tool descriptions/server instructions, poisoned agent session history, and encrypted/encoded instruction payloads), PII-minimized alerts, least privilege, hash-only default, no payment-card data in chat; never upload family/friends attachments; VT upload only for clearly suspicious unknown senders
  • skill Check domain age use this when a domain looks new, re-registered, or is impersonating a known brand and you need registration date, registrar, or archive history for a fraud or phishing check
  • skill Email attachment triage use this when an email has attachments that need a full safety pass — headers, body, readability/private-stub check, static PowerShell/hex/dropper tells, hash-only lookup by default; VirusTotal file upload only for clearly suspicious non-contact mail; severity Red/Yellow/Green; fail-closed
  • skill PowerShell obfuscation detection tells use this when statically scanning scripts/Office/attachments for attacker PowerShell obfuscation patterns (detection only; never execute samples)
  • skill Read email body use this when reading an email body (text, links, attachments) for phishing or fraud tells — display text vs href first, then DNS/reputation; hidden AI-targeted prompt injection; after or with header analysis
  • skill Read email headers use this when reading email headers to detect spoofing or manipulation, or when building a baseline of normal mail from known associates
  • skill Scan attachment hash use this when an email attachment needs a malware check: default is local hash + hash-only lookup; never upload family/friends; VT upload only for clearly suspicious unknown senders; static PowerShell -w / hex-separator / APPDATA VBS dropper tells; never run samples
  • skill Security API keys use this when this security bot needs an API key — secure input, store under secrets, rate-limit caps, rotate redundant APIs, never paste in chat; VT file upload only for clearly suspicious non-contact attachments
  • skill Threat intel sources use this when choosing threat-intel and owner toolkit sources for DNS, domain age, hashes, mail diagnostics, Shodan IP exposure, or US government feeds
  • skill Verify calendar appointment use this when creating or updating a calendar appointment — after writing the event, read it back and verify title, date, time, street-address location, and reminders before telling the user it is done
  • skill Verify download hash use this when a file is being downloaded from GitHub or another public repository (or any untrusted public source) and must be hash-checked before opening, installing, or executing
  • skill Verify link DNS use this when a link comes from another person, site, bot, app, or email and you need to check it is not listed as malicious before opening it
  • memory Memory entries · 32 items

Dependencies

No plugin dependencies are declared.

Attachments

No separate attachments are published for this record.